Send mail to:  mgnet@cs.yale.edu             for the digests or bakeoff
               mgnet-requests@cs.yale.edu    for comments or help
 
Anonymous ftp repository:  www.mgnet.org (128.163.209.19)
 
Current editor:  Craig Douglas douglas-craig@cs.yale.edu
 

WWW Sites:  http://www.mgnet.org or
            http://casper.cs.yale.edu/mgnet/www/mgnet.html or
            http://www.cerfacs.fr/~douglas/mgnet.html or
            http://phase.hpcc.jp/mirrors/mgnet or
            http://www.tat.physik.uni-tuebingen.de/~mgnet

Today's editor:  Craig Douglas (douglas-craig@cs.yale.edu)

Volume 11, Number 9 (approximately September 30, 2001)

Today's topics:

     Open Position in Darmstadt (Computational Bioelectromagetics)
     2 Open Positions at SMU
     Preprint: Preconditioning a Mixed Discont. FEM for Radiation Diffusion
     Preprint: An Unstructured Multigrid Method Based on Geometric Smoothness
     Preprint: Introduction to Robust Multigrid Technique
     Ph.D. Thesis: Stefan Reitzinger
     ETNA, Volume 12
     Workshop on Current and Future Trends in Numerical PDE's, Feb. 8-9, 2002
     UCLA MOV2001 Workshop: Dec. 3--5, 2001 
     Japanese Mirror's New URL
     Code Query

-------------------------------------------------------

Date: Tue, 04 Sep 2001 12:58:03 +0200
From: Markus Clemens 
Subject: Open Position in Darmstadt (Computational Bioelectromagetics)

The Computational Electromagnetics Laboratory (www.TEMF.de) at
the Darmstadt University of Technology invites applications for an open

                Phd/Postdoc research position.

The position involves a research project supported by the DFG (Deutsche
Forschungsgemeinschaft) centered around the numerical simulation of
current density distributions in high resolution 3D anatomy models of
human bodies. Typical applications are the exposure to slowly-varying
ambient electromagnetic fields of high intensity and their impact on
pacemakers. The project involves the solution of very large systems of
equations arising from real-valued stationary current formulations or
complex-valued eddy current formulations, for which multigrid schemes
have to be used in a distributed computational environment.

The Computational Electromagnetics Laboratory is continuously focused
on the development of methods and algorithms for the numerical simula-
tion of electromagnetic fields and their application to real world
technical problems. With its staff, consisting of about 30 researchers
at phd- and postdoc-level and its state-of-the-art research facilities,
it offers an excellent and creative environment for high-profile re-
search in the field of Computational Electromagnetics. It is the origin
of the so-called "Darmstadt School" of Computational Electromagnetics
involving the Finite Integration Method, the canonical discretization
method for Maxwell's Equations of electrodynamics.

For this research project we are looking for outstanding university
graduates with a specialization in Applied Mathematics/Scientific Compu-
tation, Computational Engineering, Electrical Engineering, Computer
Science or Physics.
Programming skills and the ability to work in a team, as well as a
strong interest in the field of scientific computing are an advantage
for this position.

Applications accompanied by the usual documents (curriculum vitae with
date of birth, diplomas, list of publications, ...) should be send to

Prof. Dr.-Ing. Thomas Weiland

Dept. Electrical Engineering and Information Technology
Laboratory for Computational Electromagnetics (TEMF)
Schlossgartenstrasse 8
64289 Darmstadt, Germany

Email: weiland@temf.tu-darmstadt.de
URL:   http://www.temf.de

-------------------------------------------------------

Date: Fri, 28 Sep 2001 16:50:44 -0500 (CDT)
From: zchen@post.cis.smu.edu (Zhangxin Chen)
Subject: 2 Open Positions at SMU

Southern Methodist University
Dedman College

Department of Mathematics

Applications are invited for two positions at either the senior level
(tenured) or junior level (tenure-track), to begin in the fall semester of
2002.  Applicants must have a strong commitment to teaching at all levels and
provide evidence of outstanding research.  The Department of Mathematics has
an active doctoral program in computational and applied mathematics with
twelve of the fifteen present faculty conducting research in these areas.
Current research includes numerical analysis of differential equations,
dynamical systems, bifurcation theory, finite element methods, perturbation
methods, and mathematical software with applications to areas such as
nonlinear optics, lasers, solidification, vortex dynamics, reservoir
simulation, pattern formation, and nonlinear waves.

To apply, send a letter of application with a curriculum vitae, a list of
publications, and a research and teaching statement to:  The Faculty Search
Committee, Department of Mathematics, Southern Methodist University, P.O.  Box
750156, Dallas, Texas 75275-0156.  Applicants must also arrange for three
letters of recommendation to be forwarded to the Faculty Search Committee.

The committee will begin its review of the applications on or about January
14, 2002.  To ensure full consideration for the positions, the application
must be postmarked on or before January 14, 2002, but the committee will
continue to accept applications until the positions are filled.  The committee
will notify applicants of its employment decision after the positions are
filled.

SMU will not discriminate on the basis of race, color, religion, national
origin, sex, age, disability or veteran status.  SMU is also committed to
nondiscrimination on the basis of sexual orientation.

Visit the department's home page at http://www.smu.edu/math for more
information.  The Search Committee can be contacted by sending e-mail to
mathsearch@mail.smu.edu.  [Tel: (214) 768-2506; Fax: (214) 768-2355]

-------------------------------------------------------

Date: Mon, 20 Aug 2001 12:50:40 -0400 (EDT)
From: benzi@mathcs.emory.edu
Subject: Preprint: Preconditioning a Mixed Discont. FEM for Radiation Diffusion

Preconditioning a Mixed Discontinuous Finite Element Method for Radiation
Diffusion
 
James S. Warsa, Michele Benzi, Todd Wareing, and Jim Morel
 
We propose a multilevel preconditioning strategy for the iterative solution of
large sparse linear systems arising from a finite element discretization of
the radiation diffusion equations.  In particular, these equations are solved
using a mixed finite element scheme in order to make the discretization
discontinuous, which is imposed by the application in which the diffusion
equation will be embedded.  The essence of the preconditioner is to use a
continuous discretization of the original, elliptic diffusion equation for
preconditioning the discontinuous equations.  We have found that this
preconditioner is very effective and makes the iterative solution of the
discontinuous diffusion equations practical for large problems.  This approach
should be applicable to discontinuous discretizations of other elliptic
equations.  We show how our preconditioner is developed and applied to
radiation diffusion problems on unstructured, tetrahedral meshes and show
numerical results that illustrate its effectiveness.

    Editor's Note: See http://www.mgnet.org/mgnet-papers.html
    -------------  or http://www.mathcs.emory.edu/~benzi/Web_papers/pubs.html

-------------------------------------------------------

Date: Thu, 20 Sep 2001 10:06:35 -0700 (PDT)
From: "Edmond Chow" 
Subject: Preprint: An Unstructured Multigrid Method Based on Geometric Smoothness

  An Unstructured Multigrid Method Based on Geometric Smoothness

                          Edmond Chow
            Center for Applied Scientific Computing
             Lawrence Livermore National Laboratory
              L-560, Box 808, Livermore, CA 94551
                        echow@llnl.gov

                           Abstract

For non-M-matrices, this paper proposes an unstructured multigrid method that
only attempts to interpolate in the directions of geometrical smoothness.
These directions are determined by analyzing samples of algebraically smooth
error, e.  Neighboring grid points i and j are called smoothly coupled if e_i
and e_j are consistently nearby in value.  In addition, these differences may
be used to define interpolation weights.  These new ideas may be incorporated
into the algebraic multigrid method.  Test results show that the new method
can have much lower grid and operator complexities compared to AMG, leading to
lower solve timings.

This work was performed under the auspices of the U.S.  Department of Energy
by University of California Lawrence Livermore National Laboratory under
contract No.  W-7405-Eng-48.

    Editor's Note: See http://www.mgnet.org/mgnet-papers.html
    -------------

-------------------------------------------------------

Date: Thu, 30 Aug 2001 11:04:42 +0000 (GMT)
From: "Serguei Martynenko (CIAM)" 
Subject: Preprint: Introduction to Robust Multigrid Technique

Introduction to Robust Multigrid Technique
Part A: Structured Grids

S.I. Martynenko
Central Institute of Aviation Motors
Moscow, Russia
martyn_s@mail.ru

ABSTRACT

Robust Multigrid Technique is intended to be a computational core of black box
software for solving physically meaningful problems on structured grids.  To
overcome problem of robustness for a sufficiently large class of the partial
differential equations, the technique consists of two parts:  ``analytical''
part (adaption of the boundary volume problems to the technique) and
``computational'' part (control volume discretization and solution of
discretized problems by original multigrid solver).  Interpolation and
pre-smoothing are eliminated from the ``computational'' part.  In addition,
the most powerful coarse grid correction strategy used in the technique makes
task of the smoother the least demanding.  Expanded robustness of the
multigrid technique is a result of adaption of equations, extremely accurate
formulation of the discrete problems on the coarse grids, original coarsening,
the most powerful coarse grid correction strategy, construction of
problem-independent transfer operators, and absence of pre-smoothing and
interpolation.

The preprint reports the essential principles of the robust multigrid
technique (Chapter 1), application to handle model problems (Chapter 2),
multigrid software (Chapter 3 and 4) and parallel implementation (Chapter 5).

    Editor's Note: See http://www.mgnet.org/mgnet-papers.html
    -------------

-------------------------------------------------------

Date: Tue, 02 Oct 2001 11:30:31 +0200
From: Reitzinger Stefan 
Subject: Ph.D. Theses

My Ph.D. theses, "Algebraic Multigrid Methods for Large Scale Finite Element
Equations," is available via the link

    http://www.trauner.at/book_detail.asp?artnr=20134361&id_titel=6131

Stefan Reitzinger
http://www.sfb013.uni-linz.ac.at/~reitz/

                                  * * * * *

Alegbraic Multigrid Methods for Large Scale Finite Element Equations
Stefan Reitzinger

Abstract

The numeral simulation of physical models is of great importance.  Instead of
performing experiments with a real life experimental setup, such an experiment
can be done virtually by computer simulation.  We use the finite element
method, which in general produces a large sparse symmetric and positive
definite system of linear equations.  In particular we concentrate on the
robust and efficient solution of the arising linear system of equations, which
is one key task in the whole numerical simulation process of many applications
appearing in science and engineering.  The present work consists in the
construction of an algebraic multigrid method for various kinds of problems.
Such methods are of interest if a finite element code does not support a
hierarchy of meshes, or if classical methods (direct or iterative linear
equation solvers) require too much memory and/or computation time.  In
addition a parallel version of the algebraic mutigrid method is constructed.
The proposed algorthms are implemented in the algebraic software package
PEBBLES.  Many numerical examples ae given in order to show the high
efficiency and flexibility of the method.

Schriftenreihe der Johannes-Kepler-Uni Linz, Reihe C, Bd.  36

1. Auflage 2001 
134 Seiten, A5, broschiert, 
ISBN 3-85487-259-3
ArtNr. 20134361
ATS 255,00
EUR 18,50

Lektorat: ekindermann@trauner.at

-------------------------------------------------------

Date: Thu, 13 Sep 2001 08:48:57 -0400 (EDT)
From: Lothar Reichel 
Subject: ETNA, Volume 12

Table of Contents, Electronic Transactions on Numerical Analysis (ETNA), vol.
12, 2000.  ETNA is available at http://etna.mcs.kent.edu and at several mirror
sites, as well as on CDROM.  Papers will be added to the volume until the end
of this year as soon as they are accepted for publication.  Presently the
following papers have been published in volume 12:

G. Meurant, Numerical experiments with algebraic multilevel preconditioners,
pp. 1-65.

H. Zhang, Numerical condition of polynomials in different forms, pp. 66-87.

M. J. Castel, V. Migallo'n, and J. Penade's, On parallel two-stage methods for
Hermitian positive definite matrices with applications to preconditioning, pp.
88-112.

R. S. Varga, Gersgorin-type eigenvalue inclusion theorems and their sharpness,
pp. 113-133.

F. B. Belgacem and S. C. Brenner, Some nonstandard finite element estimates
with applications to 3D Poisson and Signorini Problems, pp. 134-148.

S. Ehrich and A. Rathfeld, Piecewise linear wavelet collocation, approximation
of the boundary manifold, and quadrature, pp. 149-192.

J.-B. Chen and M.-Z. Qin, Multi-symplectic Fourier pseudospectral method for
the nonlinear Schrodinger equation, pp. 193-204.

B. Fischer and F. Peherstorfer, Chebyshev approximation via polynomial
mappings and the convergence behaviour of Krylov subspace methods, pp. 205-215.

A. A. Dubrulle, Retooling the method of block conjugate gradients, pp. 216-233.

------------------------------

Date: Sun, 30 Sep 2001 10:23:00 -0400 (EDT)
From: Craig Douglas 
Subject: Workshop on Current and Future Trends in Numerical PDE's, Feb. 8-9, 2002

CURRENT AND FUTURE TRENDS IN NUMERICAL PDE's:
Where is the field, and where is it going?

Friday and Saturday, Feb 8-9, 2002
Texas Institute for Computational and Applied Mathematics (TICAM)
The University of Texas at Austin.  Austin, Texas, USA

    URL: http://www.ticam.utexas.edu/~arbogast/jim.html

The field of computational PDE's is undergoing several paradigm shifts due to
the advent of powerful computers.

  - It has prompted the development of new types of algorithms, like domain
    decomposition methods for elliptic problems and the highly parallelizable
    discontinuous Galerkin methods for non-linear hyperbolic problems.
  - It has rendered the idea of strict approximation error control a plausible
    possibility.  This has resulted in the development of new continuous
    dependence results for linear and non-linear PDEs based on which adaptive
    strategies are being devised to monitor and control the error.
  - It has also been a catalyst for the devising and application of numerical
    schemes based on multi-resolution analysis.
  - It has enhanced the range of physical phenomena that can be modeled by
    PDEs and then approximated.  This is the case, for example, of models that
    take into account physical phenomena taking place in different scales and
    of models devised to capture stochastic mechanisms.

The goal of the conference is to gather different communities working in
computational PDEs and expose participants to some of the main current trends
in the field.  This meeting will challenge its participants to confront the
many aspects of the task of approximating the solutions of mathematically
sophisticated models, and it will expose them to some of the tools developed
in the different sub-communities of computational PDE's.

-------------------------------------------------------

Date: Thu, 20 Sep 2001 12:32:27 -0700 (PDT)
From: "Joseph R. Shinnerl" 
Subject: UCLA MOV2001 Workshop: Dec. 3--5, 2001 

The UCLA IPAM Workshop on Multilevel Optimization in VLSICAD has been
rescheduled for Dec. 3-5, 2001.  We again invite you to attend.  Please visit

           http://www.ipam.ucla.edu/programs/mov2001/

for more information and online registration.

-Joe Shinnerl and Jason Cong

                                  * * * * *

Editor's Note:  The talks scheduled include

  Achi Brandt (The Weizmann Institute of Science)
  Multiscale Optimization Strategies 

  Jason Cong (UCLA CS)
  Computational Challenges in Gigascale VLSICAD

  Ding-zhu Du (University of Minnesota CS&E)
  Guillotine Cut in Approximation Algorithms

  Stephan Hartmann (TU Berlin)
  New Complexity Issues and Algorithms on Channel and Switchbox Routing

  Bruce Hendrickson (Sandia National Labs)
  A Survey of Multilevel Combinatorial Methods in Scientific Computing 

  George Karypis (University of Minnesota CS&E)
  Multilevel Algorithms for Hypergraph Partitioning:  Single and Multiple
  Constraints

  Michael Lewis (College of William & Mary,  Mathematics)
  Levels of Resolution and the Optimization of Systems Governed by Differential
  Equations

  John Lillis (University of Illinois Chicago)
  Topics in Middle-Down Placement

  Stephen Nash (George Mason University, Systems Eng. & Op. Res.)
  Multigrid Algorithms for Discretized Optimization Problems 

  Rob A. Rutenbar (Carnegie Mellon University)
  Hierarchical Synthesis for Industrial-Scale Analog Intellectual Property

  Majid Sarrafzadeh (UCLA CS)
  SPS Project: Strategically Reconfigurable Systems

  Lieven Vandenberghe (UCLA EE)
  Applications of Convex Optimization in Power and Ground Network Design and
  Wire Sizing

  Chandu Visweswariah (IBM Research)
  Optimization Challenges in Transistor Sizing

  Chris Walshaw (University of Greenwich,  Computing and Math. Sci.)

  Multilevel Refinement for Combinatorial Optimisation Problems 

  Jacob White (MIT  EECS)
  Multiscale Algorithms for Electrical Peformance Analysis of Interconnect

  Gabriel Wittum (University of Iowa, Management Sciences)
  Filtering Algebraic Multigrid

  Martin D.F. Wong (University of Texas at Austin, CS)
  The Floorplan Design Problem

-------------------------------------------------------

Date: Fri, 14 Sep 2001 23:59:54 +0900 (JST)
From: NISHIDA Akira 
Subject: Japanese Mirror's New URL

Thank you for your kind permission for mirroring your site.  Due to the recent
integration of national laboratories in Japan, the Electrotechnical Laboraty,
where our site http://phase.etl.go.jp/ locates, has become a division of AIST,
the National Institute of Advanced Industrial Science and Technology.

We will appreciate if you could use the following URL as the new location of
our mirror:

    http://phase.hpcc.jp/mirrors/mgnet/

Thank you very much for your cooperation. 

Yours faithfully, 
Akira Nishida

--
NISHIDA Akira                       
Department of Computer Science, 
The University of Tokyo
7-3-1 Hongo, Bunkyo-ku, Tokyo 
113-0033 JAPAN
TEL +81-3-5841-4076
FAX +81-3-3818-1073
E-mail: nishida@is.s.u-tokyo.ac.jp

-------------------------------------------------------

Date: Thu, 20 Sep 2001 13:14:18 +0900 (KST)
From: "R. Ganesh" 
Subject: Code Query

I am a research student at POSTECH korea.

I am looking for a 2D xy Poisson solver (cartesian) which can handle circular
boundary (say conducting).  i.e., \partial^2 \varphi/\partial x^2 + y-part =
f(x,y) with varphi=0 on a circular boundary x^2 + y^2 =R^2.  in fact a uniform
grid spaced xy solver would do.

This is because I would like to avoid the 1/r problems faced in poisson
solutions in regular cylindrical polar co-ordinates at origin.

Do you have some suggestions as to availability of such code?

with regards
ganesh

    Editor's Note: If your code solves his problem, please send him email
    -------------  directly.

R. Ganesh                               e-mail : ganu@kitty.postech.ac.kr
Plasma Application  Modeling Group.     Phone  : (Off)  82-54-279-8086
Department of Electrical Engineering.   Fax    : (Off)  82-54-279-2903
Pohang University of Science and        Phone  : (Home) 82-54-279-3963       
Technology (POSTECH).                   
San-31 Hyoja, Pohang 790-784, S. Korea.

------------------------------

End of MGNet Digest
**************************